Summary of runlevel settings in init.d scripts in the Archive

Petter Reinholdtsen pere at hungry.com
Mon Aug 31 09:49:15 UTC 2009


[Sven Joachim]
>> *     2 # default-stop: 0
>
>One of these is initscripts' halt, who is the other suspect?

The other is hpoj, reported in #544346.

> Many of those 112 scripts started in runlevel S (and some of the
> others) may not need to ever be stopped because they do not start a
> daemon -- sudo comes to mind as an example.  However, 77 scripts
> which stop in runlevels 0 and 6 but not in 1 looks like a lot of
> crap.

Yes.  On the other hand, runlevel 1 is broken in so many ways already
both with the legacy boot ordering and the dependency based one, that
these errors do not make much of a difference. :/

>> Probably he flagged them as it doesn't make sense to start during s and the
>> other runlevels?
>
> Oh yes, now I see that "s 2 3 4 5" is also (correctly) flagged as
> suspicious.  Two notorious examples of these are portmap and nfs-common.

Very few packages start daemons from rcS.d/, and any package doing so
should be investigated to make sure the setting is correct.  The only
one I know about are udev, portmap and nfs-common, which are needed
for devices and NFS mounting respectively. :)

> Which it does not do currently, instead it silently ignores the
> arguments given and only uses the LSB header information.

Patches welcome.  I agree that update-rc.d should check and report
inconsistencies between the LSB header and the arguments used by
update-rc.d, at least until the legacy boot ordering API is dropped,
but never found time to implement it. :)

Happy hacking,
-- 
Petter Reinholdtsen




More information about the initscripts-ng-devel mailing list